Single-photon avalanche diodes and advanced digital circuits for improved biomedical imaging
Recent advances in biomedical imaging include the enhancement of image contrast, 3D sectioning capability, and compatibility with specialized imaging modes such as fluorescence lifetime imaging (FLIM).1–3 Compared with other imaging methods, FLIM offers the highest image contrast because it measures the lifetime of the fluorescence, rather than just its intensity or wavelength characteristics. The contrasting fluorescence lifetime attributes can then enable the observer to discriminate between regions, such as identifying healthy and diseased tissue for cancer detection. In conventional FLIM, a discrete single-photon detector, typically based on photomultiplier tube (PMT) technology, enables the acquisition of a single focal spot.4 This focal spot is then raster-scanned across the field of view to form an image. This approach, however, requires sequential scanning—pixel by pixel—and thus results in a slow image acquisition rate. In many instances, the image acquisition is critical to the application, e.g., in live cell imaging and in high-content screening for drug discovery. Consequently, there is a great need for high-speed imaging instruments that would enable next-generation biomedical imaging applications.
In current imaging techniques, high-resolution charge-coupled devices or CMOS image sensors are used. Although these imagers have high spatial resolution, their critical limitation is their poor temporal resolution. They are thus limited only to the acquisition of fluorescence intensity images. The fluorescence lifetimes of most biological samples are in the nanosecond range. The measurement of fluorescence light on such a short timescale, therefore, requires photodetectors with both very high (single-photon) sensitivity and very high (sub-nanosecond) timing accuracy. To meet these requirements, solid-state single-photon detectors—such as single-photon avalanche diodes (SPADs)5, 6—offer a promising solution. SPADs are smaller than PMT alternatives, and are also more robust, consume significantly less power, and are less expensive.
We have previously proposed that single-photon detectors, fabricated using CMOS technology,7 are the ideal sensors for emerging biomedical imaging applications (e.g., which require both single-photon sensitivity and high timing resolution). The main benefits of fabricating such devices with CMOS technology are the lower power and the faster performances that can be achieved. These benefits can be realized with lower overall system cost and with improved system miniaturization.8
In our more recent work, we have developed solid-state photon detectors that exploit standard digital CMOS technology with fast timing electronics (e.g., 130nm CMOS). We have also built a comprehensive and analytical model in a hardware description language to simulate SPAD behavior. This model is compatible with mainstream commercial design and simulation tools.9 Biasing the SPAD above its breakdown voltage (i.e., in the Geiger mode) makes it ultrasensitive to light, so that a single photon can trigger the detection. Our SPAD delivers <100ps timing resolution (the timing uncertainty between successive photon detection events). In addition, a low dark counting rate (which represents the noise level in the absence of incoming photons), of down to 1kHz, enables the detection of very weak fluorescence signals.10
We have also conducted laboratory tests to demonstrate the capability of our SPAD for accurately measuring fast fluorescence decays. A schematic representation of our experimental setup, in which we use a time-to-digital converter (TDC), is shown in Figure 1. With this setup we successfully measured the short fluorescence lifetimes of two fluorophores—Rhodamine 6G and a ruby crystal—which have fluorescence lifetimes of about 4ns and 3ms, respectively.
In biomedical imaging applications, timing information is critical. For example, in positron emission tomography (PET), time-of-flight information can be used to improve the signal-to-noise ratio of reconstructed images. We have therefore designed and implemented an advanced and compact TDC, with the use of a standard 130nm CMOS process.11 With our device we can measure the time interval between two successive photon detection events and produce a digital output that represents the time difference. This time interval can either be the interval between two coincident events in PET imaging, or the time interval between an excitation laser pulse and the SPAD's response in FLIM. The SPAD converts the optical triggering to electrical signals and feeds them to the TDC. Our TDC features a time resolution of 7.3ps, a power consumption of 1.2mW, a 0.03mm2 area, and it enables good linearity. We have integrated the SPAD and TDC in a prototype demonstration, and have conducted time-of-flight measurements, in which we achieved a 442ps ±81ps full-width at half maximum coincidence-timing resolution. When we integrated these components in the time-of-flight PET imaging system, we achieved improvements to the signal-to-noise ratio (by a factor of 2.5) and to the effective sensitivity (by a factor of 6.3).12
In summary, we have demonstrated that single-photon avalanche diode and time-to-digital converter technologies—based on standard digital CMOS devices—can deliver high speed, ultrasensitivity, and good timing resolution. They are therefore ideal candidates for improving biomedical imaging systems. Furthermore, as these technologies are fully compatible with standard CMOS processes, they offer reduced fabrication costs, smaller feature sizes, and lower power consumption compared with existing technologies. Our future work will include developing arrays of SPADs and TDCs on the same silicon substrate to build silicon photomultipliers. These may then be used in biomedical imaging and sensing applications that require both high photon sensitivity and time-stamping capability.